capacity planning

COMP9334 Capacity Planning for Computer Systems and Networks Week 2B: Queues with Poisson arrivals COMP9334 1 Pre-lecture exercise: Where is Felix? (Page 1) • Youhavetwoboxes:Box1andBox2,aswellasacatcalledFelix • Thetwoboxesareconnectedbyatunnel • Felixlikestohideinsidetheseboxesandtravelsbetweenthemusingthe tunnel. • Felixisaveryfastcatsotheprobabilityoffindinghiminthetunneliszero • YouknowFelixisinoneoftheboxesbutyoudon’tknowwhichone 12 T1, 2021 COMP9334 2 Pre-lecture exercise: Where is Felix? (Page 2) • Notation: • Prob[A]=probabilitythateventAoccurs • Prob[A|B]=probabilitythateventAoccursgiveneventB • You do

COMP9334 Capacity Planning for Computer Systems and Networks Week 2A: Operational Analysis (2). Workload Characterisation COMP9334 1 Last lecture • Modellingacomputersystemasaqueueingnetwork • Operationalanalysisonqueueingnetworks • Wehavederivedtheseoperationallaws • UtilisationlawU(j)=X(j)S(j) • ForcedflowlawX(j)=V(j)X(0) • ServicedemandlawD(j)=V(j)S(j)=U(j)/X(0) • Little’slawN=XR T1,2021 COMP9334 2 This lecture • Operational analysis (Continued) • Usingoperationallawfor • Performance analysis • Bottleneck analysis • Workload characterisation • Poissonprocessanditsproperties

COMP9334 Capacity Planning for Computer Systems and Networks Week 7A_1: Queueing disciplines COMP9334 1 Queuing disciplines Arrivals Departures • We have focused on first-come first-serve (FCFS) queues so far • However, sometimes you may want to give some jobs a higher priority than others • Priority queues can be classified as • Non-preemptive • Preemptiveresume

COMP9334 Capacity Planning for Computer Systems and Networks Week 4A: Non-markovian queueing models. Processor sharing. COMP9334 1 Week 3A: Queues with Poisson arrivals (1) • Single-server M/M/1 Exponential inter-arrivals (l) Arrivals Departures Exponential service time (μ) • By using a Markov chain, we can show that the mean response time is: T1,2021 COMP9334 2 Week